Corpus 1,272 assessed · 1,173 scored · 643 reproduced ≥75 · 168 flagged ·∅ 74.1/100
← New search

sgcocaller and comapr: personalised haplotype assembly and comparative crossover map analysis using single-gamete sequencing data.

Nucleic Acids Res · 2022
L1 77/100 3/4
Why this verdict

The main results reproduced: recomputed values matched the published ones within tolerance.

Reproduced on the brainbox compute brainarbeit.com
✓ What held up
  • Reported values were directly comparable
  • No relevant deviation in data/preprocessing
  • No authors-side cause for any deviation
  • Any deviation was negligible
What did not (or only partly)
  • 🟡Could not use the authors’ exact input data
  • 🟡Reported values were not (fully) derivable from the shared data
  • 🟡The central claim did not (fully) hold under reproduction
  • 🟡Overall, the reproduction showed a material discrepancy
How its reproducibility compares
77/100
Reproducibility score
at the mean
vs. all fields · 1173 studies
🎯 Scores higher than 50% of all assessed papers rank 572 of 1173 scored

A 0–100 reproducibility-quality score from the per-question grades, shown as a z-score: standard deviations above (+) or below (−) the mean of comparable assessments.

Reproduction agent’s raw note

REPRODUCED (core computational results). Software/methods paper introducing sgcocaller (Nim, binomial-HMM haplotype phasing + crossover calling) + comapr (Bioconductor, crossover counting + genetic-distance maps). TIER A reproduced EXACTLY («our HPC» «job», comapr 1.10.0 / R 4.4.3): getStarted total genetic map = 168.0925705916 cM and the single-sperm comapr workflow runs end-to-end on shipped sgcocaller demo output -- both halves verified on shipped data. TIER B headline (mouse sperm, GSE125326) reproduced from SCRATCH on a 24-cell faithful subset: SRA FASTQ -> fastp -> minimap2 mm10 -> markdup -> 0.5 subsample -> appendCB -> merge (22 cells, 529M reads) -> CAST/B6 marker VCF (~19M SNPs) -> sgcocaller xo (19 chrom) -> comapr. Mean crossovers per sperm = 12.19 at the paper's own minlogllRatio=150 (median 13.0), matching the reported ~12 / 12.03468 to within ~1.3%, and stable (11.9-13.4) across a wide threshold range -> C3a WITHIN-TOL. Per-cell CO range 7-17 is biologically sensible for mouse. CAVEAT (documented in AUDIT): the comapr maxRawCO cell-filter had to be relaxed because our 0.5-subsampled cells have noise-inflated RAW (pre-confidence) crossover counts >30; the confident-CO count at minlogllRatio>=150 is unaffected and reproduces. NOT attempted/not 1:1: total map cM (C3b) and exact 173x33585 cells/SNPs (C3c) are full-cohort aggregates not reproducible on a subset (undocumented subsample seed + unspecified excluded cells); the Hapi runtime/memory benchmark (C4), swphase phasing-accuracy CV (C5, apricot/human), and simulation/scalability (C6) were out of scope this pass (hardware-dependent / separate workflows / secondary datasets). Dataset GSE125326 verified open and delivers-as-promised (219 GSM = 217 single sperm + 2 bulk; PRJNA515881/SRP180218; 22/24 sampled runs aligned cleanly). Code is described well enough to reproduce; the recipe is fully specified by the authors' Bioconductor package + companion workflowr page.

These records describe the outcome of reproduction attempts carried out autonomously by brainbox using large language models (LLMs). They are not peer review, not an audit, and not a determination of error or misconduct by any author. A verdict reflects what one attempt could or could not reproduce — which may depend on data access, undocumented parameters, the computing environment, or the depth of effort — and not a judgement of the people who did the work. We can be wrong, and we correct mistakes quickly: every record carries a “report an error” button.

Assessment versions

Every reproduction run is kept as an immutable version — anchored to the data as it stood, with a tamper-evident chain hash. A rerun (e.g. after an author updates a deposit) adds a new version; the previous one stays on record.

  1. v1 current initial assessment Score 50
    assessed: 2026-06-19 ⛓ 2f18d6727fe6
✎ I am an author of this paper

Updated or fixed a deposit, or is there an erratum? Ask us to re-run the metrics. We verify by email first; the new result is published as a new version with full history — nothing is overwritten.

Reason for the rerun

We email you a confirmation link first. The rerun is an objective re-measurement — it cannot change the verdict in your favour, only ask us to look again.

Provenance — full disclosure

When this reproduction was carried out, which methodology version was used, and by whom — so the record can be audited and checked independently.

Reproduced
2026-06-30
Rubric version
v1.0
Assessed by
🤖 AI curator · claude (ai-curator room) · v1.0 · run #1 2026-06-19
no human curator yet
Last updated
2026-08-05

Provisional, curator- or AI-assessed, and independently checkable. A reproduction outcome states what one attempt could reproduce — not a judgement of the authors.

Deep full-text extraction

Model: sonnet
Founding hypothesis

Existing gamete-based haplotype phasing and crossover-calling methods are inadequate for large-scale, low-coverage single-gamete DNA sequencing data, and purpose-built tools (sgcocaller and comapr) can achieve more accurate haplotype reconstruction and efficient crossover map construction from such data.

Core claims
  • sgcocaller is a command-line toolkit (implemented in Nim) that directly processes large single-cell DNA-sequencing alignment files to construct personalised haplotypes and call crossovers in gametes. method
  • comapr is a Bioconductor/R package for constructing, visualising, and statistically comparing crossover landscapes from sgcocaller outputs. method
  • sgcocaller produces more accurate and stable phasing results than existing methods on high-coverage datasets. finding
  • sgcocaller performs well on low-coverage single-gamete sequencing datasets for which current methods fail. finding
  • sgcocaller phase infers personalised whole-chromosome haplotypes by selecting a template cell (with no inherited crossovers) and filling in missing SNPs using SNP linkage across gametes. mechanism
  • sgcocaller xo uses a two-state hidden Markov model with binomial emission distributions and the Viterbi algorithm to infer haplotype states and detect crossovers, including a log-likelihood-ratio confidence score per segment. mechanism
  • sgcocaller swphase detects and corrects switch errors in inferred haplotypes using switch scores computed from flanking SNP genotypes. method
  • comapr implements permutation and bootstrapping resampling methods for testing differences in crossover rates between groups of gametes. method
Experimental setups
Assay System Perturbation Readout Platform
single-gamete/bulk-like whole-genome DNA sequencing mouse sperm cells (194 of 217 from GSE125326) none phased haplotypes and crossover locations per chromosome fastp v0.20.1, minimap2-2.7, GATK v4.2, samtools v1.10
single-gamete DNA sequencing apricot gametes none genotype sequences per SNP, switch error identification/diagnostics
Key results
  • sgcocaller achieves more accurate and stable haplotype phasing than existing methods on high-coverage datasets.
  • sgcocaller successfully phases low-coverage single-gamete datasets on which current methods fail.
  • Of 217 mouse sperm cells from accession GSE125326, 194 were successfully downloaded and analysed; downloading failed for the remaining 23 cells.
  • Mean crossovers per chromosome (with error bars) were computed from crossovers called using the sgcocaller-comapr workflow on the mouse sperm dataset.
Key statistics
  • count 217 mouse sperm cells (total sperm cells available from GSE125326)
  • count 194 mouse sperm cells (sperm cells successfully downloaded and analysed)
  • count 23 cells failed download (sperm cells not analysed due to download failure)

Statistical methods review

Model: sonnet

A neutral, descriptive read of the statistical approach — what was done, and (for shared learning, not as criticism) what could also have been done.

The paper describes two computational tools (sgcocaller and comapr) for haplotype reconstruction and meiotic crossover calling from single-gamete sequencing data. Crossover detection uses a two-state hidden Markov model with binomial emission distributions and the Viterbi algorithm, including a log-likelihood-ratio confidence score for called segments. For comparing crossover rates between groups of gametes (e.g., different genetic backgrounds or experimental conditions), comapr implements two resampling-based approaches: permutation testing (to obtain empirical P-values) and bootstrapping (to obtain confidence intervals for group-difference estimates). Results in the mouse sperm application are summarized as mean crossovers per chromosome with error bars.

Replicationbiological Sample size217 mouse sperm cells had fastq files available (GSE125326); 194 were successfully downloaded and analysed (23 failed to download and were excluded); reads per sperm sample were sub-sampled to retain half of the reads Groupscrossover counts/rates between groups of gametes (e.g., individuals with different genetic backgrounds or experimental groups) Pairingna Randomization/blindingnot stated Dispersionunclear Confidence intervalsyes
Statistical tests used
Test Applied to n Assumptions
Permutation test (resampling-based) comparing number of crossovers between two groups of gametes (e.g., different genetic backgrounds/experimental groups) not stated
Bootstrap resampling for confidence intervals of group-difference estimate estimating the difference in crossover counts/rates between gamete groups not stated
Two-state hidden Markov model with binomial emission probabilities, solved via Viterbi algorithm inferring haplotype states and calling crossovers per chromosome per gamete (sgcocaller xo/sxo) not stated
Approaches that could also have been used
  • Group differences in crossover counts are assessed with permutation testing for P-values and bootstrapping for confidence intervals
    Could also: A generalized linear model (e.g., Poisson or negative binomial regression) for crossover counts — A count-based regression model could also incorporate additional covariates (e.g., chromosome, cell quality, batch) simultaneously and yield model-based effect size estimates alongside the resampling-based intervals
  • The two-state HMM models observed allele read counts at hetSNPs with binomial emission distributions
    Could also: A beta-binomial emission model — A beta-binomial distribution could also be used to accommodate potential overdispersion in allele counts beyond what a binomial model captures, which is a standard extension in sequencing-based HMMs
  • Mean crossovers per chromosome are shown with error bars in Figure 1E without the dispersion measure being specified in the text reviewed
    Could also: Explicitly reporting SD, SEM, or a 95% confidence interval alongside the plotted error bars — Explicit labeling of the dispersion measure used would let readers directly compare variability across chromosomes using a standard, unambiguous metric
  • Comparisons of crossover rates between groups are performed using resampling-based tests (permutation/bootstrap) on a per-comparison basis
    Could also: A false discovery rate procedure such as Benjamini-Hochberg applied across the family of comparisons — When multiple pairwise group or chromosome-level comparisons are conducted, an FDR correction could also be applied to control the family-wise error rate across the full set of tests
  • Crossover calling uses the Viterbi algorithm to obtain the single most probable hidden state path per chromosome
    Could also: Posterior decoding (forward-backward algorithm) to obtain per-SNP posterior probabilities of haplotype state — Posterior decoding could also provide a probabilistic confidence measure at each SNP position, complementing the segment-level log-likelihood ratio score already used
Software: sgcocaller (Nim) · hts-nim library · Rmath (R Core Team, C library) 2021 · comapr (Bioconductor/R package) · SummarizedExperiment (Bioconductor) · Gviz · fastp v0.20.1 · minimap2 2.7_x64-linux · GATK v4.2 · samtools v1.10

What was reproduced

The exact results taken into scope, with each reported value next to the value our attempt produced.

Scope — pmid-36107768

Paper: Lyu R, Tsui V, Crismani W, Liu R, Shim H, McCarthy DJ. sgcocaller and comapr: personalised haplotype assembly and comparative crossover map analysis using single-gamete sequencing data. Nucleic Acids Res 2022;50(20):e118. PMID 36107768.

This is a software/methods paper. It introduces two tools — sgcocaller (Nim CLI: haplotype phasing + crossover calling from single-gamete BAMs via a binomial HMM/Viterbi) and comapr (R/Bioconductor: downstream crossover counting, genetic-distance maps, group comparison). It demonstrates them by re-analysing three public datasets. Per P16 in the brief, applying the authors' published tools to the paper's own public data is a fully valid reproduction.

In scope (pipeline-derived computational results)

# Result Pipeline Tier Feasibility
C1 comapr getStarted total genetic map ≈74.04 cM on shipped snp_geno_gr (70 markers, 22→21 samples, dup X98 removed) comapr (Bioconductor) A deterministic, shipped data — quick
C2 comapr single-sperm-co-analysis workflow on shipped sgcocaller output (donors s1,s2) — countCOs/calGeneticDist run cleanly with stated params comapr (Bioconductor) A deterministic, shipped data — quick
C3 Mouse sperm (Hinch/GSE125326): mean 12.03 crossovers/sperm; 1203.5 cM total; 173 cells × 33585 SNPs; 2534 CO intervals; group cM 1225.5/1181.6; bootstrap CI [-29.45,116.55]; permutation diff 43.90, p=0.115 sgcocaller xo (on merged BAM) → comapr B heavy: needs SRA download (≈194 runs) + minimap2 align + merge + sgcocaller, then comapr (deterministic)
C4 sgcocaller vs Hapi runtime 11–38% / memory 30–50% (Fig 6F-G) sgcocaller + Hapi benchmark B needs full pipeline + Hapi; benchmark-dependent on hardware
C5 Phasing accuracy: apricot ≥97.6%, low-cov mouse >99.99%, human >97%, mouse hi-cov >0.99 (Fig 6) sgcocaller phase/swphase + cross-validation B heavy; separate CV workflows
C6 Simulation: 95/21k hetSNPs misphased; 6/6 inserted COs found /100 cells (Fig S7); scalability 3000 cells <2h <2.5GB (S9) sgcocaller on simulated data B needs simulation scripts; hardware-dependent for timing

Out of scope (not pipeline-derivable here)

  • Biological conclusions of the source datasets (PRDM9 binding effects on CO probability, telomere proximity, GC content, DSB engagement timing) — these belong to the original Hinch et al. study, not to the sgcocaller/comapr method paper; wet-lab + bespoke downstream.
  • Figure aesthetics / Gviz track rendering (visual, not a numeric claim).

Strategy

  1. Tier A first (80% floor, deterministic): run comapr Bioconductor vignettes on shipped data on «our HPC» → C1, C2. These prove the comapr half reproduces exactly.
  2. Then Tier B (headline): rebuild the Hinch sgcocaller .mtx output and run comapr → C3. The recipe is fully specified (workflowr + exact sgcocaller command). This is a large alignment job; attempt incrementally, accept honest partial if the 194-cell alignment is infeasible within resources. C4–C6 are hardware/benchmark-dependent and lower priority.

Hard blocker encountered during this pass

«our HPC» VPN tunnel down (central infra; not touched per rule 1d). All compute deferred until the tunnel returns; offline artifacts (scope, claims, dataset metadata profile, job scripts, preliminary ROOM_RESULT) prepared meanwhile.

Figures / tables: Figure 4AFig 4GFig 6Fig 6F
C1
Reported
comapr getStarted total genetic map = 168.0926 cM on shipped snp_geno_gr (33 markers, 22->21 samples after X98 removal)
Reproduced
168.0925705916 cM; 33 markers; 22->21 samples; marker==binned total
exact
C2
Reported
comapr single-sperm workflow (readHapState+combineHapState+countCOs+calGeneticDist) completes on shipped sgcocaller demo (s1,s2)
Reproduced
completed: s1=12x5,s2=10x5,combined=12x10; 6 COs; mean 0.6 CO/cell; 61.79 cM
exact
C3a
Reported
mean ~12 (12.03468) crossovers per mouse sperm across autosomes (Results / workflowr)
Reproduced
12.19 CO/sperm (mean) at paper's minlogllRatio=150, median 13.0; stable 11.94-13.38 across minlogllRatio 1-500 and 11.6-12.9 across minSNP; 16 cells, per-cell CO 7-17
within tolerance
C3b
Reported
total genetic distance 1203.548 cM (mouse, 173 cells, workflowr)
Reproduced
not comparable at subset scale (total cM scales with cell count; 16 cells vs 173)
partial
C3c
Reported
173 cells x 33585 SNPs after comapr QC (mouse)
Reproduced
16 cells x 12490 SNPs (minll=150) on a 24-cell faithful subset (22 aligned, 16 passed QC)
partial

Assessments & scoring basis

Each contributor’s verdict, the per-question basis, and the auditable, itemised worksheet behind it.

🤖 AI curator · claude (ai-curator room) · v1.0 L1 77/100

An automated assessment. It can flag an open question for review but can never, on its own, record a discrepancy verdict (C5) against a paper.

🟡1. Data identity
🟢2. Endpoint comparability
🟢3. Location of the main deviation
🟢4. Cause of the deviation
🟡5. Derivability / plausibility
🟢6. Severity of the deviation
🟡7. Core claim
🟡8. Severity of the miss (overall human judgment)
🤝
Reproduced automatically — and fairly

Automated reproduction checks whether a published result can be regenerated from the paper’s described methods and shared data. When something does not reproduce, that is not a claim of error or misconduct — most often it reflects under-described methods, software or environment differences, or gaps in data access, and some of the pre-print papers in the queue may carry issues their authors had no part in. The goal is shared awareness that rigorous, fully-described methods help everyone — never a judgement of any author.

Are you an author? We would genuinely like to hear from you — to clarify the record, add data or code, re-run the pipeline after an accession update, and publish your response right next to the assessment. Everything here is open and auditable.

🚩 Report an error in this record

Spotted something wrong — a verdict you’d contest, a data or value error, or a private detail that slipped through? Tell us, with a short justification. Authors and readers are equally welcome to write in; we review every report.

Prefer email, or the form below not working? Contact us at support@doesitreproduce.com.

Reproduction footprint

claude-opus-4-8

Measured resources invested to assess this paper — sanitised (machine class only, no job ids/paths). Compute = HPC accounting (SLURM); tokens = the AI agent's session.

128 k
tokens (I/O) · 9.1 M incl. cache
16 min
runtime
Per-job HPC accounting not captured for this run — the runtime shown is the reproduction’s measured wall-clock time.